Keyhole plasma arc welding is a complicated thermal process and has great application potential in joining important metal structures. It is of great significance to sense and monitor the keyhole geometry and its surrounding temperature distribution for seeking deep insight into the process mechanisms. In this study, an optical sensor and an infrared camera are simultaneously used to capture both the images of keyhole and its surrounding thermal fields during plasma arc welding process. Special filtering measures are taken to eliminate any interference and ensure the imaging quality. Based on appropriate imaging processing and calibration, the information from two sensors are integrated and coupled, and the shapes and sizes of both the keyhole and the weld pool are determined under different welding conditions. The observation results lay solid foundation for controlling the keyhole and weld pool dynamics, optimizing the process parameters, and validating the numerical models of thermo-physical behaviors in keyhole plasma arc welding. (C) 2013 Elsevier Ltd. All rights reserved.